Mapping desertification susceptibility in the Hodna watershed Algeria, using a remote sensing-based logistic regression approach
Abstract: This study aims to identify regions prone to desertification in the Hodna basin, which covers a vast area of the vulnerable Algerian Steppes. Thirteen factors were selected as independent variables: geomorphological factors (wind and water erosion, Topographic Position Index (TPI), Drainage Density, Slope, Aspect, Elevation), environmental factors (Normalized Difference Vegetation Index (NDVI) and Evapotranspiration), soil factors (Land surface Temperature (LST) and Normalized Difference Salinity Index (NDSI), and socio-economic factors (the Human Influence Index (HII) and the Livestock. Due to the lack of previous desertification data, NDVI anomalies served as the dependent variable. all of the variables were mainly derived using remote sensing techniques and a logistic regression model was applied for analysis in the R environment for desertification susceptibility mapping, displaying stable predictive power with a Receiver Operating Characteristic (ROC) curve of 0.76. Unlike the Northern part of the Mediterranean, where water erosion is predominant, wind erosion and soil salinization emerged as key factors in this study, while the socio-economic factors had less influence than anticipated. The resulting map shows that 45.4% of the basin is highly to very highly susceptible to desertification, provinding essential data for targeted intervention strategies.